Classification vs Regression
How should be handle Classification vs Regression Problems in the user facing api. I see 3 options:
- Kaustubh Approach: as in Kauthub's Matlab code. Two functions: One for Classification, one for Regression
- Current Approach: using one function with an argument deciding whether it is a Classification or Regression problem. The Model is then chosen appropriately. An auto mode is possible, where we try to detect which problem it is by checking the target.
- Sklearn Approach: One function, but the user is proving which kind of problem it is, by using another model.
Which do you think is the best approach for users? Have I missed smth. ?
PS: This has an effect on the Backend API:
Currently, we provide a regression and classification version for each model if possible.
They can then be accessed by using the name and the problem e.g.
- available_modes['svm']['class'] => SVC
- available_modes['svm']['reg'] => SVR
Edited by s.hamdan